This workshop consists of an introduction to the DFTB [1] formalism and its linear-scaling, massively parallel extensions, developed recently in collaborations with our group: fragment molecular orbital ([2] DFTB (FMO-DFTB) [3] and divide-and-conquer DFTB (DC-DFTB) [4]. I will highlight strengths and weaknesses of each approach, and in the hands-on application examples the user will develop a feeling for scalability and computational speed in the context of materials science problems.
There is no workshop fee, but registration is required. Please register using the link to your right.
Hours
Friday, April 8 9am-12pm, 1pm-4pm
Monday, April 11 9am-12pm, 1pm-4pm
The workshop is split into four 3-hour sections:
- In the morning session of the first day, the formalism of DFTB, FMO-DFTB, and DC-DFTB is introduced.
- In the afternoon of the same day, straightforward DFTB/MD simulations of fullerene collisions and polymer dynamics are studied in practical examples run on a local Linux cluster, highlighting the problem of the cubic scaling in DFTB. The results from MD simulations will be analyzed using JMol and/or MOLDEN software that workshop participants install on their own laptops.
- On the second day, in the morning session we will learn input preparation and perform FMO-DFTB/MD simulations of the same polymers on the computer system. This will allow us to evaluate the obtained speedup and scalability on calculations with 1, 2, and 4 CPU cores.
- In the afternoon session, the basics of input preparation for DC-DFTB is introduced, and the results for sample applications to the simulation of processes occurring at solid/gas interfaces are shown. Furthermore, we will compare the performance of DC-DFTB to FMO-DFTB for selected systems.
1] Q. Cui, M. Elstner, "Density functional tight binding: values of semi-empirical methods in an ab initio era”, Phys. Chem. Chem. Phys. 2014, 16, 14368-14377.
[2] D. G. Fedorov, T. Nagata, K. Kitaura, "Exploring chemistry with the fragment molecular orbital method”, Phys. Chem. Chem. Phys. 2012, 14, 7562–7577.
[3] Y. Nishimoto, S. Irle, "Density-Functional Tight-Binding Combined with the Fragment Molecular Orbital Method”, J. Chem. Theory Comput. 2014, 10, 4801–4812.
[4] H. Nishizawa, H. Nishimura, S. Irle, H. Nakai, “Development of linear scaling DFTB and its massive parallelization”, to be submitted.